1. Introduction

Multiphoton microscopy (MPM) is a powerful optical imaging technique in the biomedicine field by virtue of its subcellular resolution and biochemical specificity. Two major contrast mechanisms in MPM are the two-photon excitation fluorescence (TPEF) and second harmonic generation (SHG) [1,2]. TPEF can be excited either from intrinsic tissue components (e.g., nicotinamide adenine dinucleotide (NADH), elastin, and flavins) or from exogenous fluorophores. SHG can be detected from non-centrosymmetric structures such as collagen. TPEF and SHG employ two low-energy photons simultaneously to create nonlinear excitation. To increase the probability of two-photon excitation, femtosecond laser pulses are utilized together with objective lenses of high numerical aperture for the confinement and for focusing the light in time and space, respectively. The signal intensity of TPEF and SHG depends quadratically on the excitation laser power, which enables the intrinsic localized excitation from the focal volume. In addition, the use of near-infrared light for excitation reduces the scattering and improves the imaging depth. Moreover, the signal intensity of MPM is also inversely proportional to the pulsewidth of the laser [1]. In general, shorter pulses can therefore excite MPM signals more efficiently. Therefore, MPM has several advantages such as intrinsic optical sectioning and a relatively deep penetration depth, which make MPM the method of choice, e.g., in cancer research, neuroscience, and so on [3,4].

In order to apply MPM for in vivo imaging in clinical applications, bench-top MPM microscopes need to be miniaturized into compact MPM endoscopes without reducing the overall imaging quality. A lot of attention has been devoted to the development of fiber-based MPM endoscopes for high-resolution imaging [5–16]. One primary challenge in the MPM endoscope development is the delivery of ultrashort pulses with high fidelity through optical fibers in the near infrared region [17]. Several kinds of fibers, including single mode fibers (SMF) [5–7], double cladding fibers (DCF) [8–11], and photonic crystal fibers (PCF) [12–16] have been investigated. Each type of fiber has its specific advantages and restrictions. SMF introduces severe pulse broadening due to a large amount of group velocity dispersion (GVD). SMF also suffers from excessive power dependent nonlinear effects, in particular self-phase modulation (SPM) in the small fiber core [18], which severely limits the pulse energy that can be transmitted. Therefore, specific dispersion and nonlinear pre-compensation schemes are required for SMF to deliver ultrashort pulses [7]. DCF consists of a core, an inner cladding, and an outer cladding, where femtosecond pulses can be delivered through the core as a single mode and MPM signals can be collected by the inner cladding as a multimode. Thus the advantage of DCF lies in the fact that the delivery of illumination light and collection of emitted signals can be achieved through the same fiber. Same as the SMF, the limitations of the DCF are also the severe pulse broadening and substantial nonlinear effects. Even though such pulse broadening could be compensated by prism or grating stretchers [19–21], dispersion compensation increases the complexity of the system and also limits the shortest pulse that can be achieved on the sample. Furthermore, fiber nonlinear effects are generally hard to overcome if the fiber core is small, which limits the maximum light intensity that can be delivered. PCF is characterized by a periodic array of air-hole structure that extends over the entire length of the fiber. As a sub-class of PCF, photonic bandgap fiber (PBF) relies on strict photonic bandgap created by microstructured cladding to confine light in a low-refractive-index core, which is often implemented as a hollow air-filled void in an otherwise regular hexagonal lattice structure. Hence, PBFs show relatively low dispersion and nonlinear effects as compared to SMFs and DCFs. Since the material dispersion in PBF is almost negligible, the overall dispersion is dominated by waveguide dispersion, which typically changes from normal to anomalous within the propagation band [22]. Several groups have investigated the properties of PBF-based MPM systems. Göbel et al. [23] demonstrated nearly distortion-free delivery of 170 fs pulses at the zero-dispersion wavelength (812 nm) with a pulse energy of 4.6 nJ. However, above or below the exact zero-dispersion wavelength, the output pulse is significantly broadened, e.g., to 2 ps at 795 nm. Similarly, Tai et al. [24] showed that 200 fs pulses at 790 nm were stretched up to 12.4 ps after 1 m long PBF at a 35 nm offset to the zero-dispersion wavelength (755 nm). Thus pre-compensation of fiber dispersion is still needed when using broadband ultrashort lasers or tunable lasers. Studies also indicate considerable third-order dispersion (TOD) of PBF, which greatly surpasses that of SMF at the same wavelength [22]. Thus the application of PBF has been limited to delivering pulses longer than 100 fs.

To date, delivering sub-100 fs pulses to the sample through optical fiber is still a challenge, especially for broadband ultrashort sources. Recently, a specially-designed chirped photonic crystal fiber (CPCF) reported by Skibina et al. showed almost vanishing pulse broadening, which probably answers the needs of fiber-based MPM endoscopy [25, 26].

In this paper, the CPCF originally applied in biosensor is characterized as a promising candidate for fiber delivery of femtosecond pulses in MPM endoscope applications. As a proof of concept, the CPCF is tested as the delivery fiber on a bench-top MPM microscope by replacing the free-space beam path with the CPCF delivery. The feasibility of the CPCF for MPM imaging is demonstrated on various biological samples totally based on intrinsic TPEF and SHG signals. In addition, MPM images obtained from the microscope using free-space and CPCF delivery are compared. The CPCF delivery system shows comparable image quality as the free-space microscope. Moreover, we also compare MPM images obtained with the CPCF and with a traditional PBF for light delivery, indicating a dramatic enhancement of image intensity for CPCF delivery. Consequently, it is demonstrated that the CPCF has great potential for developing MPM endoscopes that include fiber delivery of sub-100 fs pulses for enhancing the MPM signals.

2. Characterization of the chirped photonic crystal fiber

2.1 Structure of CPCF

The CPCF is a particular type of the hollow core photonic crystal fibers. As shown in Fig. 1, the CPCF consists of a 50 µm hollow core and a five-layer cladding with increasing hole sizes from the inner to the outer layer. This is a distinctive feature of CPCF compared to conventional hollow core photonic crystal fibers which have uniform hole sizes. This radial chirp in cell-size is actively introduced to manage dispersion, analogous to the principle of chirped mirrors. Therefore, light of different wavelengths is reflected at different layers of the chirped cladding in the CPCF, introducing strong group delay variations with wavelength, and effectively tailoring the dispersion profile. As described more concretely in [27], the introduction of a chirp into the structure in combination with unavoidable dimensional variation of the structure along the fiber actually serves to wash out the detrimental dispersive resonances. This unique fiber design therefore enables almost complete elimination of dispersion. Moreover, the 50 µm core size of the CPCF is sufficiently large to relax the maximum power limitation due to nonlinear effects.

 

Fig. 1 Optical micrograph of the structure of a CPCF. Scale bar is 30µm.

Download Full Size | PDF

2.2 Transmission band of CPCF

In order to determine the transmission band of the CPCF, a white-light lamp which covers a wavelength range of ~500 nm is used as the light source. The light from the white-light lamp is coupled into the CPCF of ~1 m length. The spectra of the white-light source and after passing through the CPCF are measured by a spectrometer (Spectral Products, SM442). The transmission band of CPCF is obtained by dividing the output spectrum from the CPCF with the source spectrum. In this measurement, the CPCF presents two transmission wavelength-bands around 520 nm and 790 nm, respectively, as shown in Fig. 2(a). The transmission band around 790 nm is ~130 nm wide. For MPM excitation at a center-wavelength of 800 nm, the CPCF can successfully deliver the excitation light without any significant wavelength dependent distortion. Figure 2(b) shows the spectra of a Ti:Sapphire laser (Femtolasers, Fusion PRO 400) and its output after propagating in the CPCF. The laser source has a bandwidth of ~90 nm. The spectral bandwidth is only slightly reduced from 90 nm to 85 nm after passing through the CPCF, and the spectral shape remains similar to the original laser spectrum. Therefore, by providing a wide transmission band and relying on the flat dispersion characteristics, the CPCF enables the delivery of <100 fs pulses to the specimen, which can potentially increase MPM signals. The far-field beam profile after the CPCF is Gaussian like, as shown by the insertion in Fig. 2(b).

 

Fig. 2 (a) Measurement of transmission window of ~1 m CPCF by using white light source; (b) Spectra of the laser light before and after passing through ~1 m CPCF. The insertion in (b) shows the far-field beam profile after ~1 m CPCF delivery.

Download Full Size | PDF

2.3 Dispersion of CPCF

The dispersion of CPCF is characterized by measuring the pulsewidth with an autocorrelator (Femtochrome, FR-103MN). In order to test the CPCF, light from the femtosecond laser is coupled into a ~0.5 m-long CPCF through a coupling lens (Edmund Optics, NT69-476) and the light exited from the fiber end is collimated by a collimation lens (Newport, M-5X). Figure 3 shows the measured pulsewidths in three cases. First, the pulsewidth at the laser output is measured to be ~16 fs (blue curve). Second, the pulsewidth after passing through the coupling and collimation lenses alone without the fiber is measured to be ~56 fs (red curve), in order to consider the dispersion introduced by the two lenses. Finally, the pulsewidth after the coupling lens, CPCF, and collimation lens is measured to be ~67 fs (green curve). All the pulsewidth estimations rely on the fairly conservative assumption of a Gaussian pulse shape. As we can see, the two lenses introduce a significant broadening of the laser pulse. However, the CPCF does not introduce much additional broadening. This result indicates the extremely low dispersion of CPCF in the transmission band. The small side-lobes observed in the autocorrelation trace after passing through the ~0.5 m CPCF are likely caused by the slightly multi-mode of the CPCF. The CPCF design determines its preferred guiding mode to be the fundamental mode, with slight higher-order modes which experience substantially higher losses [25]. When the fiber length is increased to ~1 m, the side-lobes disappear and the CPCF can be considered as mainly single-mode.

 

Fig. 3 Intensity autocorrelation measurements for three cases without the dispersion compensation. The blue curve shows the pulse directly at the laser output; the red line shows the laser pulse after passing through the coupling and collimation lenses; the green line shows the pulse after passing through ~0.5 m CPCF in addition to the coupling and collimation lenses.

Download Full Size | PDF

For quantitative analysis, the estimation of the dispersion of the fiber and the lenses is performed based on the relationship between pulse duration and group delay dispersion (GDD). For an originally transform-limited Gaussian pulse with the duration$\Delta t$, and assuming that second-order dispersion dominates, the output pulse duration ($\Delta t_{out}$) can be expressed by [28]

After calculating the GDD of the lenses ${\phi }_{2\_Lenses}$ and the GDD of the CPCF together with the lenses ${\phi }_{2\_\left(CPCF+Lenses\right)}$, the GDD of the CPCF alone, ${\phi }_{2\_CPCF}$, can be estimated by

2.4 Loss of CPCF

Besides the dispersion characteristic, propagation loss is another important factor to be considered for MPM imaging. The Ti:Sapphire laser (Femtolasers, Fusion PRO 400) is used as the light source for the loss measurement. This laser has a center wavelength of ~780 nm and a bandwidth of ~90 nm. This wavelength range is typical for MPM imaging and it also fits in the high transmission efficiency band of the CPCF. In order to determine the propagation loss, the output powers after passing through CPCFs with different lengths are measured respectively, from which the propagation loss of CPCF is obtained under the assumption of a fixed coupling efficiency. This approach results in an estimated loss of$~5.1\pm 2dB/m$. The relatively higher propagation loss compared to conventional PBFs is presumed by its unique guiding mechanism, so called as quasi-guiding, in which the mode propagating in the air-core unavoidably leaks into the cladding layer. Nevertheless, the power loss can be compensated by increasing the laser power without increasing the complexity of the system.

In addition to the propagation loss, extra power loss caused by fiber bending is also measured. In general, physical bending of the fiber can break the confinement of the core-mode via deformation of the fiber cross-section. Since a certain degree of bending is likely to happen in MPM fiber-based endoscopy, the critical bending radius of CPCF is investigated by measuring the resulting power loss as a function of the bend radius. In this experiment, the CPCF is bent into a half-loop at various radii and the corresponding output powers are recorded. As shown in Fig. 4, when the bending radius reaches ~5.5 cm, the output power after the CPCF drops dramatically, indicating that the critical radius of the CPCF is ~5.5 cm. Therefore, for CPCF, a bending radius larger than 7 cm is recommended to avoid extra optical power loss.

 

Fig. 4 Measurement of bending loss of CPCF.

Download Full Size | PDF

3. Implementation of CPCF in MPM imaging

3.1 MPM imaging system setup

The experimental setup for MPM imaging with fiber-delivered pulses is shown in Fig. 5. The Ti:Sapphire laser (Femtolasers, Fusion PRO 400) with a center wavelength of 780 nm, full width at half maximum (FWHM) of 90 nm is used as the light source. The laser pulse duration is ~16 fs and average output power is 320 mW. For pre-compensating the overall system dispersion, a folded prism pair [20] is employed right after the laser output. Here, it should be stressed that the pre-compensation unit is mainly used to optimize the dispersion from the lenses such as the coupling, collimation, beam expansion, and high NA objective lenses where most of the system dispersion arises from.

 

Fig. 5 Fiber-delivered MPM system configuration. Dotted box shows the dispersion pre-compensation unit and dash-dotted box shows the MPM imaging sub-system; M, mirror; P, prism; ND, variable neutral density filter; FL, Focusing lens; CL, collimation lens; Obj, objective lens; F, filter; PMT, photomultiplier tube. In this setup, CPCF and PBF are bent in ~90 degrees with ~13 cm bending radius.

Download Full Size | PDF

In order to demonstrate the feasibility of CPCF for MPM endoscopy application, a CPCF with ~1 m length is inserted into the beam path to deliver femtosecond pulses to an MPM sub-system. After the dispersion pre-compensation unit, the light is coupled into the CPCF by the coupling lens (Edmund Optics, NT69-476). After the CPCF, the light is collimated by the collimation lens (Newport, M-5X). A variable neutral density filter is employed before the CPCF to adjust the power entering the fiber core.

In the imaging sub-system, the CPCF-delivered beam is raster scanned by two galvanometer mirrors, expanded by two lenses, and eventually focused onto the sample by a 40 × objective lens (Olympus, LUMPLANFL N) of 0.8 NA. The TPEF and SHG signals are collected in the backward direction after a short-pass dichroic mirror (Semrock, FF670-SDi01).The TPEF and SHG signals are further separated by a second dichroic mirror (Chroma, 450DCXRU) and then detected by two photomultiplier tubes (PMTs), respectively.

3.2 MPM system dispersion management

Since the MPM signal intensity scales inversely with the pulsewidth of the excitation light, shortening the pulsewidth at the sample location can enhance the MPM intensity, especially when the power on sample is limited by photo-damage or photo-bleaching. However, a short pulse with broad bandwidth suffers severe pulse broadening due to dispersion from the optical elements in the MPM system. Therefore, the prism-pair unit is employed to pre-compensate the pulse broadening in order to achieve the shortest pulse on the sample.

The pulse duration on the sample location is monitored by the intensity autocorrelator (Femtochrome, FR-103MN). Since the intensity autocorrelator can only be applied to a collimated beam, one of the lenses in the beam expansion is relocated to a position after the objective lens to re-collimate the laser beam. This method does not introduce any new lenses and thus characterizes the pulse duration at the sample location of the MPM system [29].

Using this method, the pulse durations are measured at the sample plane before (top row) and after (bottom row) implementing dispersion pre-compensation for three pulse delivery cases, as shown in Fig. 6. Firstly, the pulsewidth of the free-space system without using any fiber is measured. As shown in Figs. 6(a) and 6(d), the pulsewidth is stretched to ~290 fs by the objective lens and other optics before applying the dispersion pre-compensation, and it is compressed back to ~27 fs after optimizing the prism-pair for dispersion pre-compensation. Secondly, one piece of CPCF with ~1 m length and the corresponding coupling and collimation lenses are added to deliver the excitation pulses, as depicted in Fig. 5. As shown in Fig. 6(b), the pulse duration of the CPCF-delivery system is broadened to ~300 fs after all the optical components and the CPCF. Figure 6(e) demonstrates that the pulse is compressed back to ~23 fs after optimizing the prism-based pre-compensation unit. Thirdly, the CPCF is replaced by a piece of commercial PBF (Thorlabs, HC-800-01) with ~0.5 m length and the pulsewidth is measured again. A shorter PBF is selected because of its severe pulse broadening. Figure 6(c) indicates that the pulses are severely stretched to 2200 fs after all the lenses and the PBF. This stretching is attributed to the broad laser bandwidth (~90 nm), the relatively large GVD far away from the zero-dispersion wavelength, and significant TOD of the PBF in the laser bandwidth window. Even after pre-compensation, the observed shortest pulsewidth is ~770 fs as shown in Fig. 6(f), which is likely caused by the strong residual TOD of HC-800-01.

 

Fig. 6 Intensity autocorrelation measurements at sample location before (top) and after (bottom) implementing pre-compensation with three pulse-delivery configurations. (a) and (d), With free-space delivery. (b) and (e), With ~1 m of CPCF fiber delivery in the system. (c) and (f), With ~0.5 m of PBF fiber delivery in the system.

Download Full Size | PDF

This measurement demonstrates that sub-30 fs pulses can be successfully delivered to the sample plane after ~1 m CPCF delivery. It also confirms that the dedicated CPCF design enables the ultrashort pulse delivery with extremely low GVD and TOD. Furthermore, the comparable pulsewidth obtained from the CPCF-delivered system and the free-space system at sample location shows that the CPCF minimizes the fiber effects for the ultrashort pulse delivery in MPM system. As a result, the dispersion management from the CPCF itself potentially simplifies the higher-order dispersion control of fiber-based MPM endoscopes and enables a significant increase of MPM signals by delivering ultrashort pulses. It should be pointed out that the small difference of the pulse durations after pre-compensation between these two systems may come from the difference of the measurement alignment.

3.3 Imaging results with CPCF delivered pulses

With sub-30 fs pulses delivered to the sample location by the CPCF fiber, the imaging capability with the fiber delivered pulses is verified on biological samples. Yew leaf and mouse tail tendon are imaged, and the corresponding results are shown in Figs. 7(a) and 7(b). In Fig. 7(a), the autofluorescence from the stomata (holes in a leaf) and papillae on the leaf surface are observed as a well-organized pattern. From Fig. 7(b), fine collagen fibrils within the collagen bundles can also be clearly resolved, which is consistent with the dominance of type I collagen in the mouse tail tendon. All the images are color coded with TPEF in red and SHG in green.

 

Fig. 7 MPM images of yew leaf (a) and mouse tail tendon (b) when excitation light is delivered by the CPCF. Images are in false color, with TPEF in red and SHG in green. Power on the sample is about 4 mW. Pixel dwell time is 10 μs for both images. The scale bar is 50 µm.

Download Full Size | PDF

In addition, MPM images are also acquired from a human skin sample using the CPCF-delivery system. Human ethics approval has been obtained from the University of British Columbia clinical research ethics board (certificate #: H96-70499). The tissue sample (right forehead skin of a 68 year old female) was obtained from the Vancouver General Hospital Skin Care Centre surgical unit. Figure 8 shows the merged TPEF/SHG images from the skin. From Fig. 8(a) to 8(f), each image is acquired at a different depth from the surface to deeper layer of the skin. Figure 8(a) shows bright fluorescence signal with no definite shape from the stratum corneum (SC) layer where keratin, one of the endogenous fluorophores, is substantially accumulated. In Fig. 8(b), large cells from the stratum granulosum (SG) layer start to be distinguishable by TPEF contrast although some parts are partially overlaid by the signals from SC. Here, the cells show bright cytoplasm but dark nuclei because NADH in the cytoplasm has autofluorescence whereas nuclei show no fluorescence. From Fig. 8(c) to 8(d), it can be observed that the cell size tends to decrease while the cell density tends to increase. In Fig. 8(e), dense and small cells (in red) are observed simultaneously with collagen fibers (in green), which presents the junction between epidermis and dermis. The deeper location in dermis layer is revealed by strong SHG from collagen and relatively weak TPEF from elastin fibers as shown in Fig. 8(f). The optical power on the skin sample is about 5 mW.

 

Fig. 8 MPM images of human skin when excitation light is delivered by CPCF. From (a) to (f), the images are acquired at different depths where the distance to the skin surface is denoted on each image. The arrows in (f) indicate the elastin fibers. Power on the sample is 5 mW. Pixel dwell time is 10 μs for all the images. The scale bar is 50 µm.

Download Full Size | PDF

3.4 Imaging comparison between CPCF delivered and free-space microscope

The quality of the MPM images acquired by the CPCF delivered pulses is compared with that acquired by free-space delivered pulses. In this experiment, while the imaging sub-system remains the same, the femtosecond pulses are delivered either by the CPCF with the coupling and collimation lenses or by free-space (removing the CPCF and the corresponding coupling and collimation lenses). The laser power on the sample is kept the same for both configurations. The gain-setting of the PMTs and the pixel dwell time remain unchanged to make a fair comparison. Fluorescent beads (Polysciences, Fluoresbrite®YG Carboxylate Microspheres) and fish scale are imaged by the two configurations, and the corresponding results are shown in Fig. 9. In Fig. 9, the top row displays images with CPCF delivery whereas the bottom row denotes images taken by the free-space configuration. Although the images are not acquired from the exactly identical locations, representative images are chosen from locations with similar features. The MPM signal intensities obtained from the two configurations are comparable. Clearly, the images acquired by the CPCF delivery show a similar high quality as those acquired by the free space microscope. Therefore, CPCF can successfully deliver ultrashort pulses with high fidelity and is capable of providing comparable MPM images as a conventional free-space multiphoton microscope.

 

Fig. 9 Comparison of MPM images of fluorescent beads and fish scale when excitation light is delivered by CPCF (top row) and free-space (bottom row). (a) and (c), TPEF images of 6 µm fluorescent beads with 0.06 mW power on the sample. (b) and (d), SHG images of fish scale with 4 mW power on the sample. Pixel dwell time remains the same as 10 μs for all the images with and without the fiber delivery. The scale bar is 50 µm.

Download Full Size | PDF

3.5 Imaging comparison between CPCF and traditional PBF delivered pulses

As PBF is commonly used as a delivery fiber in MPM endoscopy, the quality of the MPM images acquired by the CPCF delivered pulses is also compared with that acquired by PBF delivered pulses. In the setup as shown in Fig. 5, the femtosecond pulses are either delivered by the ~1 m-long CPCF or by a piece of PBF (Thorlabs, HC-800-01) with ~0.5 m length. The power loss is estimated to be ~5 dB from the CPCF and ~0.1 dB from the PBF. The optical power on the sample is kept the same for both configurations by applying different attenuations to the laser power. Figure 10 shows the comparison of the MPM images of fluorescent beads, fish scale, and human skin acquired at similar locations using either fiber for pulse delivery. The top row shows MPM images obtained from CPCF delivery, and the bottom row denotes images obtained from PBF delivery. Here the brightness of the images in the bottom row has been increased by 10 times for a discernible visualization. Figure 10 clearly reveals that the images obtained with CPCF delivery have much higher signal intensity than those obtained with conventional PBF delivery under the same excitation power on the sample.

 

Fig. 10 Comparison of MPM images of fluorescent beads, fish scale, and human skin where excitation light is delivered by CPCF (top row) and HC-800-01 (bottom row). Image brightness in the bottom row is enhanced by 10 × to make the images discernible. (a) and (d), TPEF images of 2 µm fluorescent beads with 0.1 mW excitation power. (b) and (e), SHG images of fish scale with 4 mW excitation power. (c) and (f), MPM images from dermis layer of human skin with 5 mW excitation power. The scale bar is 50 µm.

Download Full Size | PDF

To quantify the enhancement of the signal, the TPEF and SHG intensity ratios between the images acquired by the two fiber delivery systems are calculated. The average intensity of the brightest 20% of the pixels in Figs. 10(a) and 10(d) are calculated, and a TPEF intensity ratio is obtained for the fluorescent beads. Similarly, a SHG intensity ratio is obtained from Figs. 10(b) and 10(e) for the fish scale. From the skin images in Figs. 10(c) and 10(f), both TPEF and SHG intensity ratios can be obtained for the TPEF channel showing the elastin fibers and the SHG channel showing the collagen fibers, respectively. The enhancement factor is further averaged over multiple locations on the three types of samples imaged by the two fiber systems to alleviate sample inhomogeneity that may cause the variation of MPM strength. Although the origins of MPM contrasts may be different for the three types of samples, comparing CPCF to PBF, our results indicate an average enhancement factor of $12\pm 4.6$ and $50\pm 5$for TPEF and SHG, respectively. Therefore, TPEF and SHG signals are significantly enhanced by using CPCF over HC-800-01. This level of enhancement is consistent among different types of samples.

The enhancement is mainly attributed to the shorter pulse duration on the sample for the CPCF delivered system compared to the HC-800-01 delivered system. As shown in Fig. 6, the pulse duration by CPCF delivery is ~23 fs while that by PBF delivery is ~770 fs at the sample location after dispersion pre-compensation. Thus, at the sample location, the pulse duration is ~35 times shorter when delivered by the CPCF than by the HC-800-01.

The approximate 12 times enhancement of the TPEF intensity is found to be less than the ratio of the pulse durations. This may be caused by the limited absorption bandwidth of the fluorophores, which can reduce the enhancement of TPEF intensity, arising from the $1/{\tau }_p$relationship, if the laser bandwidth becomes broader than the absorption band [29]. The SHG signal is anticipated to have a higher enhancement than TPEF because SHG is not limited by the absorption band. The 50-times enhancement in SHG intensity is slightly higher than the pulsewidth ratio, which may have been caused by several reasons. First, the polarization direction of the light is different after the CPCF and HC-800-01 delivery. Since SHG intensity is sensitive to polarization, different polarization after the two fiber transmission may lead to different SHG intensities [30]. Second, errors may appear on the estimated pulse durations and their ratio because the real pulse shapes may not be exactly the same as the assumed shape (i.e., Gaussian shape).

For endoscopic MPM applications, both the excitation light delivery and the emission signal collection will need to be achieved through optical fibers. With its capability to deliver sub-30 fs pulses, CPCF can be a good candidate for delivering ultrashort pulses in MPM endoscopy to excite MPM signals more efficiently with shorter pulses. The relatively high loss in the visible range has limited the CPCF application as the simultaneous collection fiber. Nevertheless, MPM signal collection can be achieved by using a separate multimode fiber. A possible implementation of the CPCF in the MPM endoscopy is to employ the CPCF as the delivery fiber for increasing the nonlinear excitation efficiency and a multimode fiber with large core for maximizing the collection efficiency of MPM signals.

4. Conclusions

The CPCF concept has been demonstrated for potential applications in MPM endoscopy. The properties of CPCF are characterized in detail, and its pros and cons for MPM imaging are discussed. The CPCF is found to be suitable for delivering ultrashort pulses for MPM imaging due to its broad bandwidth and low dispersion. Using a ~1 m-length CPCF, sub-30 fs pulses are successfully delivered to the sample location in a MPM system. To the best of our knowledge, this is the shortest pulse that has been achieved in a MPM system that includes an optical fiber for pulse delivery. This favorable property has significant meaning in enhancing the MPM intensity by using shorter pluses. Moreover, the performance of CPCF applied in MPM imaging is evaluated and validated by label-free imaging of various biological samples, demonstrating the potential of CPCF for MPM endoscopy. In addition, the MPM images obtained by CPCF delivered pulses are found to be comparable to those acquired by free-space microscopy. Furthermore, by comparing CPCF with commonly used PBF, a significant increase of MPM image intensity is achieved, where ~12 and 50 times improvements are revealed for TPEF and SHG, respectively. Therefore, CPCF appears as a promising choice for fiber-based multiphoton endoscopy to increase the MPM excitation efficiency by delivering ultrashort pulses, which is critical for in vivo imaging where the optical power on sample and the integration time are limited.